Strengthening Password Authentication using Keystroke Dynamics and Smartphone Sensors

Abstract

Presently, a password authentication is a weak point for security in the authentication scheme because a password is easy to be stolen and a user may ignore the security by using a simple password, which is easy to remember or using the same password for all accounts. From the related works, basic keystroke dynamics features, i.e. key hold time, latency time, and interkey time, were studied on a smartphone. The results showed the weak aspect to use only basic keystroke dynamics for authentication on the phone. In this research, the study of smartphone sensors combining with keystroke dynamics is proposed to strengthen the password authentication, called a biometric authentication. New features are key hold pressure, finger area, and accelerometer sensors. The classification techniques in this work are Naïve Bayes, k Nearest Neighbors (kNN), and Random Forest. The classification accuracy percentage and equal error rate (EER) are used for measuring the performance of the features and classifiers. From the results, Random Forest gives the best performance and if all smartphone sensors and keystroke dynamics are used as features, the best accuracy percentage is at 97.90% and EER is at 5.1%.